A coherent Raman scattering imaging methodology to visualize and quantify pharmaceutical compounds within the skin is described. This paper describes skin tissue preparation (human and mouse) and topical formulation application, image acquisition to quantify spatiotemporal concentration profiles, and preliminary pharmacokinetic analysis to assess topical drug delivery.
Cutaneous pharmacokinetics (cPK) after topical formulation application has been a research area of particular interest for regulatory and drug development scientists to mechanistically understand topical bioavailability (BA). Semi-invasive techniques, such as tape-stripping, dermal microdialysis, or dermal open-flow microperfusion, all quantify macroscale cPK. While these techniques have provided vast cPK knowledge, the community lacks a mechanistic understanding of active pharmaceutical ingredient (API) penetration and permeation at the cellular level.
One noninvasive approach to address microscale cPK is coherent Raman scattering imaging (CRI), which selectively targets intrinsic molecular vibrations without the need for extrinsic labels or chemical modification. CRI has two main methods-coherent anti-Stokes Raman scattering (CARS) and stimulated Raman scattering (SRS)-that enable sensitive and selective quantification of APIs or inactive ingredients. CARS is typically utilized to derive structural skin information or visualize chemical contrast. In contrast, the SRS signal, which is linear with molecular concentration, is used to quantify APIs or inactive ingredients within skin stratifications.
Although mouse tissue has commonly been utilized for cPK with CRI, topical BA and bioequivalence (BE) must ultimately be assessed in human tissue before regulatory approval. This paper presents a methodology to prepare and image ex vivo skin to be used in quantitative pharmacokinetic CRI studies in the evaluation of topical BA and BE. This methodology enables reliable and reproducible API quantification within human and mouse skin over time. The concentrations within lipid-rich and lipid-poor compartments, as well as total API concentration over time are quantified; these are utilized for estimates of micro- and macroscale BA and, potentially, BE.
Methodologies to assess cPK after topical drug product application have expanded from classical in vitro permeation testing (IVPT) studies1,2,3,4,5 and tape-stripping6,7,8 to additional methodologies such as open-flow microperfusion or dermal microdialysis9,10,11,12,13,14. There are potentially various local sites of therapeutic action depending on the disease of interest. Hence, there may be a corresponding number of methodologies to assess the rate and extent to which an API gets to the intended local site of action. While each of the aforementioned methodologies has its advantages, the major disadvantage is the lack of microscale cPK information (i.e., the inability to visualize where the API goes and how it permeates).
One noninvasive methodology of interest to estimate topical BA and BE is CRI, which can be broken down into two imaging modalities: CARS and SRS microscopy. These coherent Raman methods enable chemically specific imaging of molecules via nonlinear Raman effects. In CRI, two laser pulse trains are focused and scanned within a sample; the difference in energy between the laser frequencies is set to target vibrational modes specific to the chemical structures of interest. As CRI processes are nonlinear, a signal is only generated at the microscope focus, allowing for three-dimensional pharmacokinetic tomographic imaging of the tissue. In the context of cPK, CARS has been used to obtain tissue structural information, such as the location of lipid-rich skin structures15. In contrast, SRS has been utilized to quantify molecular concentration as its signal is linear with concentration. For ex vivo skin specimens, it is advantageous to carry out CARS in the epi-direction16 and SRS in transmission mode17. Therefore, tissue samples that are thin will allow for SRS signal detection and quantification.
As a model tissue, the nude mouse ear presents several advantages with minor drawbacks. One advantage is that the tissue is already ~200-300 µm in thickness and does not require further sample preparation. In addition, several skin stratifications are seen by axially focusing through one field of view (e.g., stratum corneum, sebaceous glands (SGs), adipocytes, and subcutaneous fat)16,18. This allows for preliminary preclinical estimation of cutaneous permeation pathways and topical BA estimates before moving to human skin samples. However, the nude mouse model presents limitations such as difficulty in extrapolation to in vivo scenarios due to differences in skin structure19. While the nude mouse ear is an excellent model to obtain preliminary results, the human skin model is the gold standard. Although there have been various commentaries on the suitability and applicability of frozen human skin to accurately recapitulate in vivo permeation kinetics20,21,22, the use of frozen human skin is an accepted method for the evaluation of in vitro API permeation kinetics23,24,25. This protocol visualizes various skin layers in mouse and human skin while quantifying API concentrations within lipid-rich and lipid-poor structures.
While CRI has been utilized across numerous fields to specifically visualize compounds within tissues, there have been limited efforts investigating the cPK of topically applied drug products. To evaluate the topical BA/BE of topical products using CRI, it is necessary to first have a standardized protocol in place to make accurate comparisons. Previous efforts using CRI for drug delivery to the skin have demonstrated variability within the data. As this is a relatively new application of CRI, establishing a protocol is critical to obtain reliable results18,26,27. This approach only targets one specific wavenumber in the biological silent region of the Raman spectrum. However, most APIs and inactive ingredients have Raman shifts within the fingerprint region. This has previously posed challenges due to the inherent signal arising from the tissue in the fingerprint region. Recent laser and computational advances have removed this barrier, which can also be utilized in combination with the approach presented here28. This approach presented here allows for the quantification of an API, which has a Raman shift in the silent region (2,000-2,300 cm-1). This is not limited to the physiochemical properties of the drug, which might be the case for some previously mentioned cPK monitoring methodologies29.
The protocol must reduce sample-to-sample variability in skin thickness for various preparations, as thick human skin samples will produce minimal signal after drug product application due to light scattering by the thick sample. A goal of this manuscript is to present a tissue preparation methodology that assures reproducible imaging standards. In addition, the CRI system is setup as described to reduce potential sources of error as well as minimize signal-to-noise. However, this paper will not discuss the guiding principles and technical merits of the CRI microscope as this has been previously covered30. Finally, the extensive data analysis procedure is explored to allow for interpretation of the results to determine an experiment's success or failure.
The use of nude mouse ear tissue was approved by Massachusetts General Hospital Institutional Animal Care and Use Committee (IACUC), while the use of human skin tissue was approved by the Massachusetts General Hospital Institutional Review Board (IRB). According to IACUC protocols, freshly euthanized mice were obtained from collaborators with nude mice colonies. Human tissue was procured from elective abdominoplasty procedures at Massachusetts General Hospital via an approved protocol. In addition, specific tissue types other than abdominal skin were acquired via a body donation authority, also through an IRB-approved protocol.
1. Preparation of tissue
Figure 1: Images of ideal thickness for imaging mouse and human skin. (A) Mouse ear skin held up to light, which can visibly let light through. (B) Ideal human skin held up to light after preparation. Please click here to view a larger version of this figure.
2. Laser and microscope setup
Figure 2: Schematic layout for coherent Raman laser imaging path. Beams are independently conditioned for spot size and matched via time delay stage to generate coherent Raman scattering in samples for the desired tuning frequency. Please click here to view a larger version of this figure.
3. Lipid imaging
Figure 3: Example skin depths obtained using SRS. The top set of images are from nude mouse ear skin depicting the following: (A) stratum corneum, (B) sebaceous glands, (C) adipocytes, (D) subcutaneous fat. The bottom set of images are obtained from human skin depicting the following: (E) stratum corneum, (F) papillary dermis, and (G) a sebaceous gland. Scale bars = 100 µm. Both mouse and human skin images were acquired using a 20x objective at 1024 pixels x 1024 pixels; the human SG was taken at 512 x 512 pixels. Abbreviations: SRS = stimulated Raman scattering; SG = sebaceous gland. Please click here to view a larger version of this figure.
4. Application of topical formulation
5. Experimental setup for drug quantification
Figure 4: Tissue movement in nude mouse ear skin demonstrated by visualizing sebaceous glands. Example of limited tissue movement is depicted in A and B, while substantial tissue movement is depicted in C and D. (A) shows the sebaceous glands at the time of formulation application and (B) the same depth at 120 min after application. (C) Mouse sebaceous glands at the time of formulation application and (D) 120 min after formulation application; the sebaceous glands are barely visible, which is an indication that this experiment was not measuring the uptake into the sebaceous glands for the entire experimental duration. Scale bars = 100 µm. Images are 1024 pixels x 1024 pixels. Please click here to view a larger version of this figure.
6. Data analysis
Figure 5: Intensity vs. time profiles. (A) An example of flux profiles that have reached saturation and thus only a decrease in intensity is seen. Each ROI has a different flux profile to demonstrate the heterogeneity in the data that one might acquire. (B) An example of concentrations that increase after imaging has begun. Each ROI is a different field of view (indicated by the different color traces) within the same tissue of the same experiment. In addition to global concentrations, there is the ability to elucidate which local environment an API/formulation prefers as indicated by lipid-rich and lipid-poor regions. The profiles presented in A indicate that there is no absorption of drug into the tissue as the API has already permeated and begun to leave the tissue once imaging has started. However, in B, the tissue has not reached saturation, and there is still absorption of the API followed by elimination. The segmentation of images into lipid-rich and lipid-poor will aid in the elucidation of the localization of the API (or inactives) and the permeation pathways into the skin (i.e., stratum corneum). A higher concentration within the lipid-rich regions indicates that the API localizes within the lipid structure of the layer under investigation, which aids in targeted drug delivery information. Abbreviations: ROI = region of interest; API = active pharmaceutical ingredient. Please click here to view a larger version of this figure.
Imaging is considered successful if the tissue has not significantly moved in either axial (<10 μm) or lateral direction upon the completion of the experiment (Figure 4). This is an immediate indication if the SRS measurement for the API of interest is not representative of the initial depth, for which quantification is layer-specific. This is mitigated by imaging z-stacks for each XY position of interest, with the trade-off being the temporal resolution. If frozen skin is used in these studies, the penetration and permeation of the API are rapid compared to fresh skin, and minimal time between formulation application and the start of imaging is imperative. Another consideration is the objective used for the experiments. If using a 60x objective, selecting a flat surface within a single field of view (FOV) is relatively easy; however, for a 20x objective, the viewed field is much larger, and thus, a key step in the tissue preparation is ensuring uniform contact of the skin with the glass-bottom imaging dish. A flat FOV and similar depth compared to the original depth within the FOV are two keys to positive results.
The alignment of the laser, in addition to the dynamic range of the image, must be addressed with the utmost care. Misalignment of the laser pulse trains into the microscope can lead to a host of problems, including low signal levels or unevenly excited FOVs, which may give rise to low-contrast images. Another consideration is to ensure that the entire dynamic range of intensity values are utilized when acquiring images; otherwise, the imaging data are compressed, and concentration differences may be difficult to detect.
Another consideration is that skin heterogeneity may give rise to variation in the computed microscale flux profiles within the same dataset. The intensities (a proxy for concentrations) begin high and decrease over the experimental duration (Figure 5A), while other studies indicate an increase followed by a decrease in flux over the experimental duration (Figure 5B). At present, absolute concentration quantification cannot occur instantly due to the experimental setup. Thus, concentrations that decrease after application are possibly the result of a saturation within the depth of interest, and the quantification only represents the API elimination. If the dynamic range is not large enough or the skin is too thick, visual inspection will render the concentrations to appear stagnant so that no change occurs over the imaging duration. This is a function of both a suboptimal dynamic range and skin thickness, which will suggest a need to repeat the experiment.
The concentration-time profiles are then subjected to NCA of each profile to estimate the exposure, maximum flux, and time to maximum flux. Statistical analysis (Figure 6) is conducted across experimental conditions to further investigate covariates contributing to potential differences in exposures. Comparisons of the global cPK parameters will provide insight into which formulation provides a higher flux or greater exposure. In contrast, the microscale cPK parameters (i.e., the lipid-rich and lipid-poor regions) will provide insight into the local biodistribution and permeation pathways. For example, when comparing two formulations with the same API concentration and differing in the inactive ingredients, one formulation may tend to permeate through the stratum corneum via the lipid-rich region versus that of the lipid-poor region. This observation indicates that this specific formulation will "push" the permeation toward the lipid-rich regions for easier penetration and permeation.
Figure 6: Example NCA analysis of concentration-time profile from mouse ear tissue. (A) An example of tmax (time at which maximum concentration occurs) analysis between two formulations for the same API. This analysis indicates that the 2-(2-ethoxyethoxy)ethanol provides extended API permeation compared to that of the Gel formulation, regardless of the skin layer. It can also be seen that the tmax for the SC layer is longer than the SG, suggesting that the 2-(2-ethoxyethoxy)ethanol formulation is continuing to deliver API even when the imaging duration has concluded. (B) An example of total exposure analysis between the same formulation/API combination in A but at depths further into the skin. This figure is modified from18. Abbreviations: SC = stratum corneum; SG = sebaceous gland; AD = adipocyte; SCF = subcutaneous fat. Please click here to view a larger version of this figure.
Supplemental Table S1: Example dataset acquired from manual image analysis. The columns indicate the information acquired from the Data Analysis section of this manuscript (i.e., frame, area, mean, min, max, median), while additional columns have been added for use in a cPK analysis (i.e., layer, region, time_minutes). These data can be analyzed via NCA and plotted to visualize the concentration profile within the SG skin layer. Please click here to download this Table.
The evaluation of topical BA/BE is an area of research the requires a multifaceted approach as no single method can fully characterize in vivo cPK. This protocol presents a methodology for the evaluation of a topical drug product's BA/BE based on coherent Raman imaging. One of the first points that might be overlooked is how thin the skin samples must be, especially for quantitative transmission SRS imaging. If the skin is too thick (i.e., light cannot readily pass through), there is little to no signal measured by the SRS detector, and it will therefore provide poor concentration data. Care must be taken to prepare these tissue samples properly, as this can make or break an experiment. In terms of the nude mouse ear, aside from rinsing with PBS and patting it dry to remove any residual dirt on the ear, little preparation is required.
Mouse ears are typically of a thickness of several hundred µm, which is optimal for transmission SRS. Human skin samples typically are several mm thick, including subcutaneous fat, although the thickness is highly variable depending on the anatomical source of the skin tissue and the age of the donor. Thus, as much excess tissue must be removed as possible to accurately quantify lipid structures and API concentrations in the epidermis and dermis over time. If the tissue preparation step is overlooked, then the remaining steps of the experimental set-up will largely be unsuitable as the starting conditions are not optimal.
The laser/microscope setup is the next challenge or potential stumbling block. Misalignment of the laser in the beam path and ultimately into the microscope results in poor contrast and, therefore, poor imaging results. It is recommended to set up the CARS channel first, as its signal is easier to find. The pump and Stokes pulse trains must be overlapped in both time and space. The alignment of the pump laser is checked using the transmission detector within the microscope MC software when the SRS detector is flipped out of the way. While viewing the CARS channel (ALG1) in the MC software, the Stokes beam is unblocked. However, if there is no signal from the oil sample, it is first necessary to align the Stokes beam and then adjust the time overlap. It may be necessary to iterate these two adjustments until the signal is optimized. The spatial overlap of the two beams is viewed through an IR viewer on the irises, while the time delay stage (Figure 2) is adjusted to have the beams overlap in time. These two alignment steps are critical to ensure CARS signal generation.
Once a signal is apparent in the CARS channel, the SRS channel (ALG2) is the next one to set up. Potential problems for lack of signal are that the lock-in phase or gain settings are too low, or that the offset is set too high within the lock-in software. In addition, the condenser position can be adjusted to focus the transmitted light onto the photodiode and therefore optimize the SRS signal. The improper setup of the laser/microscope will lead to a lack of signal, thus decreasing concentration estimates and a lack of permeation information. The laser power of the pump and Stokes beams may be optimized for individual studies. However, it is critical that the powers of the beams are the same for each experiment. Different laser powers between replicates will give false differences in concentration, which will be due to the setup rather than the API/formulation.
Each study will require a unique dose-duration time (i.e., duration of time that the formulation is left on the skin) and must be independently investigated to quantify cutaneous API penetration/ permeation as this is formulation-dependent. Another consideration when developing a protocol is the occlusive nature of the formulation application. It is important to know whether the formulation was designed to be administered under occlusive or nonocclusive conditions. The CRI methodology presented here utilizes an inverted microscope; this means that the skin surface is face-down and under occlusive settings. An upright microscope may provide the opportunity to have nonocclusive conditions; however, the skin surface may not be flat, which would make these types of experiments challenging.
It must be acknowledged that the occlusive nature of these experiments is not the typical clinical usage; nevertheless, permeation pathways are parsed out within these studies. The CRI method presented here provides the ability to visualize and quantify microscale changes that are otherwise indistinguishable with methodologies such as dermal microdialysis, dermal open-flow microperfusion, tape-stripping, or IVPT studies. Recent developments of rapid wavenumber tuning have paved the way for the concurrent quantification of the skin structure and multiple vibration bonds outside the silent region. However, further computational methods to parse out the contribution of specific analytes from that of skin are still under development28. This is also of particular relevance for in vivo CRI studies, although the powers utilized on the benchtop in this setup (approximately 50 mW at the focus) may not be permissible for clinical use. The potential of this methodology to be translated from a lab bench to the clinic can enable investigators to quantify the drug's permeation in vivo as well as ex vivo in the same setting to develop in vitro–in vivo relationships that are crucial for the advancement in topical drug development.
The sheer amount of data acquired from one experimental run can be anywhere from 10 images per site to 70 images per site. If there are multiple sites per piece of tissue, this leads to gigabytes of information. The images themselves provide global concentration-time data and are quantified as they are, without preprocessing. However, that does not maximize the utility of CRI as local biodistribution data can be extracted in addition to permeation pathway data. The image segmentation is time-consuming but provides detailed information not possible with other methodologies. For example, it is possible to estimate the preferred penetration pathway through the stratum corneum (lipid-rich or lipid-poor), which can provide insight into which inactive ingredients might contribute to a specific pathway or if it is drug-dependent. Analysis of one experiment might take several hours to days, depending on the number of images and experimental duration. Hence, an automated approach will assist in data analysis and provide consistent annotation of lipid-rich and lipid-poor regions across skin stratifications18.
The authors have nothing to disclose.
The authors would like to thank Dr. Fotis Iliopoulos and Daniel Greenfield of the Evans' Group for their discussion and proofreading of this manuscript. In addition, the authors would like to acknowledge support from LEO Pharma. Figure 2 was created with BioRender.com.
Tissue Preparation | |||
Autoclavable Biohazard Bags | FisherBrand | 22-044562 | As refered to in text: biohazard bags https://www.fishersci.com/shop/products/fisherbrand-polyethylene-biohazard-autoclave-bags-without-sterilization-indicator-8/22044562?searchHijack=true&searchTerm= 22044562&searchType=RAPID& matchedCatNo=22044562 |
Cell Culture Buffers: Dulbecco's Phosphate-Buffered Salt Solution 1x | Corning | MT21030CV | As refered to in text: PBS https://www.fishersci.com/shop/products/corning-cellgro-cell-culture-buffers-dulbecco-s-phosphate-buffered-salt-solution-1x-8/MT21030CV?searchHijack=true&searchTerm= 21-030-cv&searchType= RAPID&matchedCatNo=21-030-cv |
Disposable Scalpels | Exel International | 14-840-00 | As refered to in text: scalpel https://www.fishersci.com/shop/products/exel-international-disposable-scalpels-3/1484000?keyword=true |
High Precision 45° Angle Broad Point Tweezers/Forceps | Fisherbrand | 12-000-132 | As refered to in text: forceps https://www.fishersci.com/shop/products/high-precision-45-angle-broad-point-tweezers-forceps/12000132#?keyword= |
Kimwipes Delicate Task Wipers, 1-Ply | Kimberly-Clark Professional Kimtech Science | 06-666 | As refered to in text: task wiper https://www.fishersci.com/shop/products/kimberly-clark-kimtech-science-kimwipes-delicate-task-wipers-7/06666 |
Parafilm M Laboratory Wrapping Film | Bemis | 13-374-12 | As refered to in text: parafilm https://www.fishersci.com/shop/products/curwood-parafilm-m-laboratory-wrapping-film-4/1337412 |
Petri Dish (35 mm x 10 mm) | Fisherbrand | FB0875711YZ | As refered to in text: small petri dish https://www.fishersci.com/shop/products/fisherbrand-petri-dishes-specialty-6/FB0875711YZ?keyword=true |
Petri Dish (60 mm x 15 mm) | Fisherbrand | FB0875713A | As refered to in text: large petri dish https://www.fishersci.com/shop/products/fisherbrand-petri-dishes-clear-lid-12/FB0875713A?keyword=true |
Surgical Scissors | Roboz | NC9411473 | As refered to in text: scissors https://www.fishersci.com/shop/products/scissors-327/NC9411473?searchHijack=true&searchTerm= RS-5915SC&searchType=RAPID& matchedCatNo=RS-5915SC |
Laser/microscope | |||
650/60 nm BrightLine single-band bandpass filter | Semrock | As refered to in text: CARS filter – CH2 vibrations (645nm/60nm filter) | |
Control box IX2-UCB | Olympus | As refered to in text: Control Box | |
D700/30m | Chroma | As refered to in text: CARS filter – deuterated band https://www.chroma.com/products/parts/d700-30m |
|
DeepSee Insight | Spectra-Physics | As refered to in text: Laser https://www.spectra-physics.com/f/insight-x3-tunable-laser |
|
Digital Handheld Optical Power and Energy Meter Console | ThorLabs | PM100D | As refered to in text: power meter https://www.thorlabs.com/newgrouppage9.cfm?objectgroup_id=3341 |
Fluoview Software | Olympus | As refered to in text: Microscope Control software | |
Frosted Microscope Slides | FisherBrand | As refered to in text: microscope slides https://www.fishersci.com/shop/products/fisherbrand-frosted-microscope-slides-4/22265446 |
|
FV1000 | Olympus | As refered to in text: Microscope | |
Incubation Chamber | Tokai Hit | GM-800 | As refered to in text: incubation chamber |
Integrating Sphere Photodiode Power Sensor | ThorLabs | S142C | As refered to in text: photodiode https://www.thorlabs.com/newgrouppage9.cfm?objectgroup_id=3341 |
Power supply FV31-PSU | Olympus | As refered to in text: Power Supply | |
Precision 4063, 80MHz Dual Channel Function Generator | BK Precision | As refered to in text: function generator | |
ProScan – Precision Microscope Automation | Prior Scientific Instruments | As refered to in text: stage controller https://www.prior.com/microscope-automation/inverted-microscope-systems/proscan-linear-stage-highest-precision-microscope-automation |
|
SecureSeal Imaging Spacers | Grace Biolabs | 654004 | As refered to in text: spacer https://gracebio.com/product/secureseal-imaging-spacers-654004/ |
SRS Detection Kit | APE | As refered to in text: SRS detector | |
UPLSAPO 20X NA:0.75 | Olympus | As refered to in text: 20X Objective https://www.olympus-lifescience.com/en/objectives/uplsapo/ |
|
Lipid/Drug Imaging | |||
35 mm Dish, No. 0 Uncoated Coverslip, 14 mm Glass Diameter | MatTek Corporation | NC9711297 | As refered to in text: Glass bottom dish https://www.fishersci.com/shop/products/glass-bottom-mircrowell-dish/nc9711297 |
Cotton-tipped applicators | FisherBrand | As refered to in text: Cotton-tipped applicator | |
Distriman Postive Displacement Pipette | Gilson | As refered to in text: Postive Displacement Pipette https://www.fishersci.com/shop/products/gilson-distriman-positive-displacement-repetitive-pipette/F164001G#?keyword= |
|
Distriman Postive Displacement Pipette Tips | Gilson | As refered to in text: Tips for pipette https://www.fishersci.com/shop/products/gilson-distritip-syringes-6/f164100g?keyword=true |
|
Data Analysis | |||
FIJI | Open-source | As refered to in text: FIJI/ImageJ https://imagej.net/software/fiji/ |
|
Jupyter-Lab | open-source | As refered to in text: JupyterLab https://jupyter.org/ |
|
Rstudio | Open-source | As refered to in text: Rstudio https://www.rstudio.com/ |